Use monotonic_time for pruning stale Sonar nodes

There's an apparent race condition in Sonar between pruning stale nodes
on `:ping` and updating the status after a notification. This primarily
happens in development for two reasons:

1. Development laptops are most prone to time warp because of system
   sleep
2. Apps only run a single node in development

Using `monotonic_time/1` instead of `system_time/1` guards against clock
drift/time warp effects.

lib/oban/sonar.ex

@@ -75,7 +75,7 @@ defmodule Oban.Sonar do
   end
 
   def handle_info({:notification, :sonar, %{"node" => node} = payload}, state) do
-    time = Map.get(payload, "time", System.system_time(:millisecond))
+    time = Map.get(payload, "time", System.monotonic_time(:millisecond))
 
     state =
       state
@@ -111,8 +111,8 @@ defmodule Oban.Sonar do
   end
 
   defp prune_stale_nodes(state) do
-    stale = System.system_time(:millisecond) - state.interval * state.stale_mult
-    nodes = Map.reject(state.nodes, fn {_, recorded} -> recorded < stale end)
+    stale = System.monotonic_time(:millisecond) + state.interval * state.stale_mult
+    nodes = Map.reject(state.nodes, fn {_, recorded} -> recorded > stale end)
 
     %{state | nodes: nodes}
   end

test/oban/sonar_test.exs

@@ -54,10 +54,10 @@ defmodule Oban.SonarTest do
 
     test "pruning stale nodes based on the last notification time" do
       name = start_supervised_oban!(notifier: Oban.Notifiers.Isolated)
-      time = System.system_time(:millisecond)
+      time = System.monotonic_time(:millisecond)
 
-      Notifier.notify(name, :sonar, %{node: "web.1", time: time - :timer.seconds(29)})
-      Notifier.notify(name, :sonar, %{node: "web.2", time: time - :timer.seconds(31)})
+      Notifier.notify(name, :sonar, %{node: "web.1", time: time + :timer.seconds(29)})
+      Notifier.notify(name, :sonar, %{node: "web.2", time: time + :timer.seconds(31)})
 
       nodes =
         name